Interface SCIP methods by adamj34 · Pull Request #1237 · scipopt/PySCIPOpt

adamj34 · 2026-07-25T12:58:03Z

Adds the following methods:
getNNodesLeft(), getNRuns(), getNReoptRuns(), addNNodes(), getDeterministicTime(), getAvgDualbound(), getMaxTotalDepth(), getNBacktracks(), getFocusNode(), getAvgLowerbound(), getFirstPrimalBound(), getLowerboundRoot(), getUpperbound(), getNObjlimLeaves()

Adds tests in tests/test_statistics.py and tests/test_node.py for migrated methods

Adds function signatures to scip.pyi for type hints

Updates CHANGELOG.md

adamj34 · 2026-07-25T13:28:16Z

@Joao-Dionisio were there some changes in the ruff config? I didn't even touch these lines in my commits.

If it's ok I can refactor it to use the more modern X | Y syntax as suggested in the error message

This is likely just a new ruff version flagging some errors. I'd just merge without caring much about this, but if you're offering to fix this, I'm more than happy to accept :)

I'll take a little bit to review, but everything seems to be in order. Can you please make sure that the optimized_model() manages to find primal solutions?

I checked that optmized_model() wasn't able to find any primal solution, so I 've increased the node limit parameter and now it finds 1.

Regarding the X | Y syntax, I only changed it in scip.pyi to pass the pipeline checks and because this syntax is already used there.

I was thinking about replacing typing.Union in the entire repo, but in general such fixes are considered unsafe for Python versions prior to 3.10 (see for reference: https://docs.astral.sh/ruff/rules/non-pep604-annotation-union/) and pyproject.toml only requires python>=3.8. Maybe there's no point in touching it until the project moves to python>=3.10?

Yes, we tend to prefer supporting as many versions as possible, and 3.10 is not thaaat old yet.

codecov · 2026-08-09T17:54:32Z

Codecov Report

❌ Patch coverage is 50.00000% with 10 lines in your changes missing coverage. Please review.
✅ Project coverage is 57.29%. Comparing base (9547994) to head (2587c2c).
⚠️ Report is 9 commits behind head on master.

Files with missing lines	Patch %	Lines
src/pyscipopt/scip.pxi	50.00%	10 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1237      +/-   ##
==========================================
- Coverage   57.91%   57.29%   -0.62%     
==========================================
  Files          26       27       +1     
  Lines        5807     5936     +129     
==========================================
+ Hits         3363     3401      +38     
- Misses       2444     2535      +91

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

Joao-Dionisio · 2026-09-01T08:15:46Z

Sorry for the delay @adamj34 ! I'm not as free in September as I was expecting to be, but I'll still manage to take a closer look.

I'm not sure what the code coverage bot is complaining about (am AFK), but it'd be nice to ensure that everything is tested.

Co-authored-by: João Dionísio <57299939+Joao-Dionisio@users.noreply.github.com>

Joao-Dionisio · 2026-09-01T18:15:51Z

+# very simple model with 2 primal solutions
+# created because getting more than 1 primal solution from random_mip_1 requres setting a large node limit, which slows down the tests
+@pytest.fixture
+def optimized_model_with_primal_solutions():


There's no guarantee that this will remain working for future SCIP versions, but I suppose the same would be true for every model.

While I'm not sure how likely it was to break in the future, I think it would be better to have something more reliable than assuming that an x number of node limit will produce y number of solutions.

I came up with choosing a simple model and providing solutions by hand, so that it's easy to reason about all possible solutions. I also added a few comments in the tests to make them clearer.

Please let me know if this approach is better.

Joao-Dionisio · 2026-09-01T21:39:37Z

+    upperbound = optimized_model_with_primal_solutions.getUpperbound()
+
+    assert isinstance(first_primal, float)
+    assert optimized_model_with_primal_solutions.isGT(first_primal, upperbound)


One is in the original space and the other in the transformed space, right?

Yes, getUpperbound() is in the transformed space and getFirstPrimalBound() is in the original space. However, I've refactored this test a bit and no longer use getUpperbound().

Joao-Dionisio · 2026-09-01T21:44:09Z

+    leaves, children, siblings = optimized_model.getOpenNodes()
+    open_nodes = leaves + children + siblings
+    manual_avg_lowerbound = 0.0
+    if len(open_nodes) > 0:


This is 0 if the model is solved to optimality, so I suppose we should assert that it isn't.

I deleted the if statement.

Joao-Dionisio · 2026-09-01T21:46:57Z

+
+
+def test_getAvgDualbound(optimized_model):
+    avg_dualbound = optimized_model.getAvgDualbound()


This has the same original vs transformed space fragility. But, I suppose it's not a big deal, it's a minimization problem, and any SCIP changes to the way it handles this will error out very loudly everywhere.

I think that for a minimization problem getAvgDualbound() and getAvgLowerbound() will be exactly the same.
For a maximization problem, one will be the negative of the other. That's why I included or in the assert statement.

We could test it similarly to test_getAvgLowerbound but there's no getDualbound() method defined on Node. This is the reason behind using a comparison to getAvgLowerbound for testing method.

adamj34 · 2026-09-03T16:07:34Z

Thank you for your suggestions and comments @Joao-Dionisio!

I believe I've addressed all of them. If there's anything that requires further improvement, then let me know.

Regarding the Codecov complaints, I double checked if all methods that I added were tested. It seems that getFocusNode was only indirectly tested in test_getAvgLowerbound, so I added a dedicated test for it in e1caacd. Beyond that, I think that all methods have an accompanying test.

PS I hope that these reviews aren't taking too much time away from your PhD 😊

adamj34 commented Jul 25, 2026

View reviewed changes

adamj34 marked this pull request as ready for review July 25, 2026 19:46

Joao-Dionisio reviewed Sep 1, 2026

View reviewed changes

adamj34 added 10 commits September 1, 2026 16:40

add getMaxTotalDepth method

e969506

add getNBacktracks and getFocusNode method

33c3ce7

add getAvgLowerbound method

1d1bcf0

add getAvgDualbound method

7f0566e

add getDeterministicTime method

d00054f

add getFirstPrimalBound method

7ca8620

add getLowerboundRoot method

9de5f77

add getUpperbound method

dbbeb9a

add getNObjlimLeaves method

d3ae10e

increase optimized_model's node_lim to get primal solutions

a39137e

adamj34 force-pushed the interface branch from 2587c2c to a39137e Compare September 1, 2026 14:44

adamj34 and others added 3 commits September 1, 2026 17:49

update CHANGELOG.md

2307b0a

Co-authored-by: João Dionísio <57299939+Joao-Dionisio@users.noreply.github.com>

update the description of getNBacktracks

9a3d450

refactor statistics tests

6ba6467

Joao-Dionisio reviewed Sep 1, 2026

View reviewed changes

adamj34 marked this pull request as draft September 1, 2026 18:58

add test for getFocusNode

e1caacd

Joao-Dionisio reviewed Sep 1, 2026

View reviewed changes

Comment thread tests/test_statistics.py Outdated

Joao-Dionisio reviewed Sep 1, 2026

View reviewed changes

Comment thread src/pyscipopt/scip.pxi Outdated

Joao-Dionisio reviewed Sep 1, 2026

View reviewed changes

adamj34 added 2 commits September 2, 2026 22:30

update getFirstPrimalBound description

8c0c6f8

improve statistics tests

5139b57

adamj34 marked this pull request as ready for review September 3, 2026 15:44



		def test_getAvgDualbound(optimized_model):
		avg_dualbound = optimized_model.getAvgDualbound()

Uh oh!

Conversation

adamj34 commented Jul 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Joao-Dionisio Sep 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

codecov Bot commented Aug 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Joao-Dionisio commented Sep 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

adamj34 Sep 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

adamj34 Sep 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

adamj34 Sep 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

adamj34 commented Sep 3, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

adamj34 commented Jul 25, 2026 •

edited

Loading

Joao-Dionisio Sep 1, 2026 •

edited

Loading

codecov Bot commented Aug 9, 2026 •

edited

Loading

Joao-Dionisio commented Sep 1, 2026 •

edited

Loading

adamj34 Sep 3, 2026 •

edited

Loading

adamj34 Sep 3, 2026 •

edited

Loading

adamj34 Sep 3, 2026 •

edited

Loading